Conference PaperPDF Available

Topology optimization of pollen filters using an adjoint solver

Authors:

Abstract

Conventional filters consist of one or more filter layers, which are either woven or composed of tangled fibers. The quality of the separation results almost only from the density of the fiber arrangement. Due to the manufacturing process, compromises between separation and pressure loss that are in the opposite relationship to each other are inevitable. The development of additive manufacturing has made great progresses in recent years. Using this technology, filter manufacturing is no longer restricted to conventional methods. At this point, the adjoint topology optimization of filters is a promising alternative opportunity of filter development. The idea behind this is to use the adjoint solver to generate geometries that have been optimized with regard to pressure loss and the filtration efficiency. The complex non-linear relationships between the separation mechanisms and the pressure drop offer the opportunity for improvement. The fluid region with the CAD model of the initial geometry is divided into volume cells. The cells at the fibers are then varied by mesh deformation until an optimum is found. The often bionic-looking optimized structures can be realized using additive manufacturing. In a further process step the structure can be coated. The new approach offers the chance to develop filters in a new way, avoiding the previous parametric development and even crossing the Pareto front in the multi-objective optimization of filtration efficiency and pressure loss. The key to topology-optimized "bionic" filters are suitable cost function(s) controlling the optimization. These cost functions have to account for the different separation mechanisms and the pressure loss. The combination of both functions was realized by an algorithm. The separation mechanisms depend on the particle size distribution, the properties of the particles and the ambient fluid, as well as the flow itself. The dominant separation mechanism can be reinforced and the part of the surface that favours an irrelevant deposition mechanism is reduced and thus its contribution to the pressure drop. The separation mechanisms are determined using initial geometries in a multi-phase microsimulation: Lagrangian simulation for the impaction and discrete element method (DEM) for interception and diffusion. The proof of concept has already been provided: In one deformation step, it was possible to optimize the initial geometry with regard to two opposing aims. The whole workflow was tested for a car pollen filter on simple initial geometries. To validate the improvement, current pollen filters were evaluated: on the one hand, a geometry was generated on the basis of a µCT scan and examined in CFD with regard to pressure loss and separation efficiency. On the other hand, samples of the filter material were tested in the test bench.
Topology optimization of pollen filters using an adjoint solver
Natalie Jüngling*, Thilo Gaugler, Jan Pospichl, Jennifer Niessner
Institute of Flow in Additively Manufactured Porous Media (ISAPS), Heilbronn
University of Applied Sciences, Max-Planck-Str. 39 74081 Heilbronn - Germany
Conventional filters consist of one or more filter layers, which are either woven or
composed of tangled fibers. The quality of the separation results almost only from the
density of the fiber arrangement. Due to the manufacturing process, compromises
between separation and pressure loss that are in the opposite relationship to each
other are inevitable.
The development of additive manufacturing has made great progresses in recent
years. Using this technology, filter manufacturing is no longer restricted to conventional
methods. At this point, the adjoint topology optimization of filters is a promising
alternative opportunity of filter development. The idea behind this is to use the adjoint
solver to generate geometries that have been optimized with regard to pressure loss
and the filtration efficiency. The complex non-linear relationships between the
separation mechanisms and the pressure drop offer the opportunity for improvement.
The fluid region with the CAD model of the initial geometry is divided into volume cells.
The cells at the fibers are then varied by mesh deformation until an optimum is found.
The often bionic-looking optimized structures can be realized using additive
manufacturing. In a further process step the structure can be coated. The new
approach offers the chance to develop filters in a new way, avoiding the previous
parametric development and even crossing the Pareto front in the multi-objective
optimization of filtration efficiency and pressure loss.
The key to topology-optimized "bionic" filters are suitable cost function(s) controlling
the optimization. These cost functions have to account for the different separation
mechanisms and the pressure loss. The combination of both functions was realized by
an algorithm. The separation mechanisms depend on the particle size distribution, the
properties of the particles and the ambient fluid, as well as the flow itself. The dominant
separation mechanism can be reinforced and the part of the surface that favours an
irrelevant deposition mechanism is reduced and thus its contribution to the pressure
drop. The separation mechanisms are determined using initial geometries in a multi-
phase microsimulation: Lagrangian simulation for the impaction and discrete element
method (DEM) for interception and diffusion.
The proof of concept has already been provided: In one deformation step, it was
possible to optimize the initial geometry with regard to two opposing aims. The whole
workflow was tested for a car pollen filter on simple initial geometries. To validate the
improvement, current pollen filters were evaluated: on the one hand, a geometry was
generated on the basis of a µCT scan and examined in CFD with regard to pressure
loss and separation efficiency. On the other hand, samples of the filter material were
tested in the test bench.
KEYWORDS
Topology Optimization, Filter development, Computational Fluid Dynamics,
Additive manufacturing, Multi-phase Microsimulation, pollen filter
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.